Implementation of GNSS/INS-Assisted Structure from Motion for a Thermal Camera Onboard a UAS

Seyyed Meghdad Hasheminasab

Seyyed Meghdad Hasheminasab Lyles School of Civil Engineering, Purdue University, West Lafayette, IN 47907-1971 USA.
Tian Zhou Lyles School of Civil Engineering, Purdue University, West Lafayette, IN 47907-1971 USA.
Ayman Habib Lyles School of Civil Engineering, Purdue University, West Lafayette, IN 47907-1971 USA.

25D

Current advances in remote sensing technologies and platforms have encouraged the adoption of thermal imaging sensors onboard Unmanned Aerial Systems (UAS) in various application domains such as precision agriculture, crash scene reconstruction, and public safety surveillance. Derivation of color-coded dense point clouds using RGB cameras on board UAS has matured and the user community has access to several commercial packages that provide accurate products at high level of automation. These packages rely on Structure from Motion (SfM) strategies for automated aerial triangulation of the acquired images. For thermal images, most of SfM strategies would not lead to a good product due to the poor texture of the available images. The main challenge for these strategies is their inability to identify sufficient number of key features in the images, which are conducive to reliable image matching. The advent of prosumer grade GNSS (Global Navigation Satellites System)/INS (Inertial Navigation System) position and orientation systems, which could be mounted onboard a UAS, could help in alleviating this problem.

This work provides a SfM strategy for using the provided position and orientation trajectory by the GNSS/INS unit to increase the reliability of the triangulation process while solving for the mounting parameters (lever arm and boresight angles relating the GNSS/INS unit to the camera). More specifically, using nominal values for the mounting parameters, the GNSS/INS trajectory information will be used to restrict the search space for the identification of conjugate features in stereo-images. A global SfM strategy is then used to refine the identified matches in the different stereo-images while solving for the mounting parameters as well as distortion parameters in the utilized camera.

The performance of the developed strategy will be tested using UAS-based images to evaluate its performance to an existing SfM approach while evaluating their comparative performance in the following cases: a) well-textured RGB imagery over an agricultural field, b) poorly-textured RGB imagery over a shoreline, and c) an image block captured by a thermal camera over an agricultural field. The comparative performance will be also evaluated using an acquired LiDAR surface model over the test areas.

14:15 Implementation of GNSS/INS-Assisted Structure from Motion for a Thermal Camera Onboard a UAS, Seyyed Meghdad Hasheminasab

January 30 @ 14:15
14:15 — 14:30 (15′)

Mineral A

Seyyed Meghdad Hasheminasab

Add to Google Calendar